2gvd.pdb has a ligand named 
which is not selected as "ligand" by Jmol 13.0.8
It is selected as "nucleic" due to ? being a nucleotide analog --though rather different
correct. The definition of "ligand" was changed in 12.2. From the documentation:
(originally "hetero and not solvent"; changed to "!(protein,nucleic,water,UREA)"
I think you just have to work with that. Nothing I can think of would be the perfect "ligand" definition.
Yes, I was aware of the ligand redefintion --and I am happy with it, I think it's better.
The issue is why this ligand should be recognized as nucleic. It is quite different from a nucleotide. I guess it has the proper atom IDs to fit into Jmol algorithm?
I assume there is no solution for this. The problem was, using standard defualt rendering of a pdb, the ligand is invisible.
Perhaps single-residue "proteins" and "nucleics" could be selected somehow and applied ball and stick instead of cartoon... Something similar to C-only proteins and P-only nucleics getting a specific treatment.
I like the idea of single-residue groups rendering in ball and stick.